Does the proctor fill in the survey for respondents?

In this administration method, respondents supply their own answers; the proctor organizes and supervises the process.

Is a proctor survey automatically anonymous?

No. Anonymity depends on identifiers and collection arrangements, not merely on completing forms in a room.

Does a high session response rate remove bias?

No. Coverage, nonresponse, wording, and the influence of the setting can still affect results.

AI Glossary / Statistics and probability / Beginner

Proctor survey

Definition

A proctor survey is a survey administered to respondents gathered in a room with a person who supervises its administration. The proctor provides structure and procedural assistance while respondents complete their own answers. The method aims to combine organized participation with confidentiality. Confidentiality and good response rates depend on the actual procedures; the presence of a proctor does not automatically guarantee either truthful answers or representative results.

What it means

The proctor manages the survey session rather than supplying respondents' answers. Typical responsibilities include explaining the instructions, distributing materials, keeping the session orderly, and clarifying procedural questions consistently. The method differs from an interviewer-administered survey in which an interviewer asks questions and records each person's responses.

Its defining feature is the supervised group setting, not a particular questionnaire topic. It can be used with paper forms or digital questionnaires. In AI-related projects, it might support structured feedback on a training session or a tool, but it is a survey-administration method rather than an AI algorithm.

How it works

Define the target population and the method for inviting respondents separately from the administration mode. Arrange a setting in which people can complete the questionnaire privately, with consistent instructions and sufficient time. Train the proctor to provide neutral procedural assistance and avoid explaining questions differently to different participants.

Specify how completed answers will be collected, who can access them, and whether identifiers are recorded. A confidential survey may retain controlled identifiers; an anonymous survey does not provide the same identifying link. Do not promise anonymity when the collection system or attendance records can connect people with answers.

After the session, distinguish invitations, attendance, returned questionnaires, and answered items. Record problems such as interruptions or inaccessible formats. Evaluate coverage, nonresponse, and possible effects of the supervised setting on responses. Organized attendance can simplify administration without resolving every source of survey error.

Three practical examples

A training provider collects private participant feedback during a supervised session immediately after an AI workshop.
A company gathers structured usability feedback from invited employees while a neutral coordinator explains the completion process.
A research team administers the same questionnaire to several groups using standardized proctor instructions and protected response collection.

Think of it this way

Think of a supervised library study session. Someone organizes the room, explains the process, and handles practical issues, but each person does the work independently. The analogy is limited: a survey should not be treated as a test with preferred answers, and participants must understand the actual privacy arrangements.

Worked example

In an original survey example, 120 eligible employees are invited. Of these, 108 attend the session and 102 return a questionnaire. A particular question receives 96 valid answers, including 72 yes responses.

Using returned questionnaires as the completion criterion, the response proportion among invitees is 102/120 = 85%. Among attendees, it is 102/108, approximately 94.44%. The valid-answer proportion for that item among returned questionnaires is 96/102, approximately 94.12%.

The yes proportion among valid item answers is 72/96 = 75%. It is not automatically the yes proportion among all 120 invitees. If all 24 invitees without a valid item answer would have said no, the full-invitee proportion would be 72/120 = 60%. If all would have said yes, it would be 96/120 = 80%.

These simple bounds assume the recorded answers are accurate. They illustrate why a strong return rate within the room does not eliminate uncertainty about people or items that remain unobserved.

Illustrative case study

Fictional case study: Cedar Learning evaluates an introductory AI course with a proctor survey. The facilitator initially plans to circulate among participants while they complete a form asking them to rate that same facilitator.

The team instead appoints a neutral coordinator, explains the collection process, and arranges private submission. It also tracks attendance separately from answer files and offers an accessible completion option. This fictional case illustrates attention to administration and perceived privacy; it does not establish that the resulting responses are unbiased or anonymous in every implementation.

Common misconceptions and mistakes

A proctor survey is not an examination and should not encourage a correct or approved response. The proctor's authority, wording, or proximity can still influence participants, particularly when the survey concerns the organization hosting the session.

Another mistake is equating confidentiality with anonymity. Controlled access to identifiable records is different from collecting records that cannot be linked to respondents. High attendance also does not prove a representative sample: people invited or able to attend may differ from the target population. Finally, a returned questionnaire may contain unanswered items, so overall completion and item response should be reported separately.

Business use

Proctor surveys can make structured feedback collection practical when participants are already assembled. They support consistent instructions and provide a clear opportunity to complete the questionnaire, which can reduce some logistical barriers.

For a business introducing AI tools, the method may help gather training or usability feedback. Plan carefully if participants are evaluating supervisors, sensitive workplace practices, or decisions that affect them. Report the sample, setting, participation figures, and privacy procedure alongside results. An AI-generated summary should preserve these limitations instead of presenting the responses as the views of everyone in the organization.

Check your understanding

Ninety-five percent of attendees complete a survey, but only half of the invited population attends. Can the attendee completion rate describe participation across all invitees?

Answer: No. The denominator differs. If these are the only completions, the invitee completion proportion is 0.95 × 0.50 = 47.5%, and the missing participants still require consideration.

Related terms

Coverage error
Questionnaire error
Survey administration
self-administered questionnaire
confidentiality
anonymity
response rate
item nonresponse
social desirability

Learn next

Compare survey administration modes, then study sampling and response-rate definitions. Learn questionnaire design, neutral instructions, and privacy-aware collection practices. Review nonresponse and measurement error before interpreting feedback as evidence about a wider population.

Measurement error

Sources

NIST/SEMATECH's glossary defines a proctor survey as a supervised group survey and describes its intended combination of confidentiality and administrative structure: https://www.itl.nist.gov/div898/handbook/glossary.htm

The denominator calculations, participation bounds, operational examples, and fictional case are original. The entry describes the survey-administration sense, not remote exam-proctoring software or a named research instrument.

Previous: Processing Bias Next: Product of experts